HN compresh 22d ago

Show HN: We matched full-context recall on ~1% of the tokens (open benchmark)

Compresh reduces context window costs by reconstructing query-aware conversation slices, addressing context degradation in LLM apps.

HN wahnfrieden 22d ago

GPT 5.6 Pro SVG output is INSANE

Clickbait headline about GPT 5.6 with no substantive content.

HN katomek 22d ago

Show HN: Exeria Charts – A native, high-performance financial charting library

Financial charting library for web applications with React support.

HN rusk 22d ago

Sideways, Not Up: Why AI Is Not a New Abstraction Layer

Opinion piece arguing AI represents sideways progress rather than new abstraction layer in software engineering.

HN fmarzochi 22d ago

EGC – your AI never re-explains your project when you switch sessions

EGC provides persistent memory across sessions for AI coding tools via local runtime, reducing context re-establishment overhead.

HN car 22d ago

Forge – Code-Quality Guardrails for AI Agents

Forge provides code-quality guardrails for AI agents via deterministic CLI/shell hooks, preventing silent standard degradation.

HN Imustaskforhelp 22d ago

The Theoretical Limit of Image Compression [video]

Stub video post about image compression with no content.

HN Saurabh_06 22d ago

Show HN: We built an open-source tool for real-time speech-to-speech translation

PolyTalk: open-source self-hosted real-time speech-to-speech translation tool using Ollama, Faster Whisper, and Piper.

HN GeorgeWoff25 22d ago

GLM 5.2 vs. Claude Opus 4.5

Comparison of Claude Opus 4.5 and GLM-5.2 frontier reasoning models on specs, pricing, benchmarks, and agent workloads.

HN ayi 22d ago

Ask HN: Anthropic banned me from using Claude Code and I don't know what to do

User discusses account bans from Anthropic's Claude Code service after using VPN, seeking solutions.

HN timothygao 22d ago

Show HN: MinLlama – Llama 3.2 inference in ~100 lines of NumPy

MinLlama is a ~100-line NumPy implementation of Llama 3.2 inference with PyTorch and Jax variants, designed for transformer research and KV cache compression.

HN avijeetsingh16 22d ago

Open source, global vs. proprietary but for US in US, which is fundable in SaaS?

Discussion about open-source Earth Memory protocol for AI agents to leverage geospatial intelligence versus SaaS business model.

HN Roselol 22d ago

Show HN: Free VIN Decoder with AI Summaries on Cloudflare Workers and NHTSA Data

VIN decoder tool using NHTSA data and AI summaries deployed on Cloudflare Workers.

HN aihatterer 22d ago

Show HN: Code Stitcher – The anti-agentic workflow

Code Stitcher tool integrates AI-generated outputs into codebases with accuracy, positioned as human-in-the-loop alternative to agentic workflows.

HN yellow-cucumber 22d ago

URL and HTML to Markdown API Optimized for LLMs

API service converting URLs and HTML to Markdown format optimized for LLM input processing.

HN javatuts 22d ago

Desktop Commander MCP

Desktop Commander MCP is a model context protocol server for code/text work and task automation, with optional dedicated app for macOS and Windows.

HN abhishekucs 22d ago

Show HN: AI Command Center

AI Command Center platform for managing team agents, connections, and knowledge with examples of sales and lead research automation.

HN shonkone 22d ago

Guidance injection: reliable instructions for local LLMs

Guidance injection technique addresses local LLM instruction drift by reliably maintaining instructions throughout long conversations.

HN gzpoffline 22d ago

Picot: Codex style GUI for the Pi coding agent

Picot is a local desktop GUI for the Pi coding agent with bundled runtime, no cloud dependency, available for macOS.

HN patriceckhart 22d ago

Show HN: Zot now has model shortcuts: jump between your favorite models

Zot is a lightweight coding agent harness written in Go that auto-downloads and installs model binaries with SHA-256 verification across OS platforms.

HN GuardCalf 22d ago

Show HN: WSL Dashboard – an open-source,low-memory, high-performance GUI manager

WSL Dashboard: Rust/Slint-based open-source GUI manager for Windows Subsystem for Linux instances.

HN toomuchtodo 22d ago

Prototyping data tools with AI, a case study: Solar and Battery Atlas

Case study on using AI to prototype and scale analytical tools, demonstrating faster iteration from research questions to working data tools.

HN jongalarraga 22d ago

Stripe pre-launch security checklist for indie SaaS

Practical Stripe payment security checklist for indie SaaS launches covering common implementation mistakes.

HN alexandroskyr 22d ago

Show HN: A private pager for your AI agent loops

Open-source tool enabling 30+ autonomous AI agents to request human input when blocked via a PWA communication channel with push notifications.

HN Floukie 23d ago

Show HN: Magpie-Search

Federated search engine that indexes local AI agent work history, enabling agents and LLMs to find and recover context after system crashes.

Ax Eric V. Strobl 23d ago

Fast Nonparametric Conditional Independence Testing via Two-Stage Regression

arXiv paper on fast nonparametric conditional independence testing for causal discovery. ML research on statistical testing and causal inference.

Ax Hiranya Garbha Kumar, Minhas Kamal, Balakrishnan Prabhakaran 23d ago

CAOA -- Completion-Assisted Object-CAD Alignment

arXiv paper on CAD model alignment in 3D scans using completion. Computer vision research for 3D reconstruction.

Ax Nicolas Mahler 23d ago

Kernel of Partition Paths: A Unified Representation for Tree Ensembles

arXiv paper on tree ensemble geometric representation and feature importance. ML theory research on decision trees.

Ax Kenneth Ge, Andre Assis 23d ago

ClayBuddy: A Framework, Evaluation, & Mitigation of Coding Agent Failures

arXiv paper systematizing coding agent failures with framework and mitigations. Research identifying critical failure modes in AI agents and safety mechanisms.

Ax Ashim Dhor, Rasel Mondal, Pin-Yu Chen 23d ago

ITNet: A Learnable Integral Transform That Subsumes Convolution, Attention, and Recurrence

arXiv paper unifying convolution, attention, and recurrence via integral transforms. Fundamental ML research on neural network architectures.

Ax Xilun Chen, Shao-Chuan Wang, Baykal Cakici, Lukasz Heldt, Lichan Hong, Raghu Keshavan, Aniruddh Nath, Li Wei, Xinyang Yi 23d ago

Token Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models

arXiv paper on integrating diverse signals into large recommendation models efficiently. ML research on transformer-based recommendation systems.

Ax Aditeya Baral, Radoslav Ralev, Iliya Sotirov Zhechev, Srijith Rajamohan, Jen Agarwal 23d ago

Closing the Calibration Gap in Semantic Caching

arXiv paper on semantic caching calibration for LLM inference. Technical research optimizing caching metrics and deployment decisions for LLM cost reduction.

Ax Jonghoon Lee, Seong Hyeon Park, Byungwoo Jeon, Minha Lee, Jinwoo Shin 23d ago

Pose6DAug: Physically Plausible Multi-view Object Swapping for Robot Data Augmentation

arXiv paper on data augmentation for robot vision-language-action policies. ML research for improving robotic manipulation training efficiency.

Ax Yusuf Salcan (Computer Vision Group, University of Freiburg, Germany, CRIION-AI Lab, Freiburg, Germany), Simon Ging (Computer Vision Group, University of Freiburg, Germany, Adaptive & Agentic AI), Robin Tibor Schirrmeister (Department of Radiology, Medical Center -- University of Freiburg, Germany), Philipp Arnold (Department of Radiology, Medical Center -- University of Freiburg, Germany), Elmar Kotter (Department of Radiology, Medical Center -- University of Freiburg, Germany), Behzad Bozorgtabar (Adaptive & Agentic AI), Thomas Brox (Computer Vision Group, University of Freiburg, Germany) 23d ago

Scalable Training of Spatially Grounded 2D Vision-Language Models for Radiology

arXiv paper on vision-language models for radiology with automatic dataset curation. ML research on domain-specific VLM training without manual annotation.

Ax Asa Shepard, Jeannie Albrecht 23d ago

Probe-and-Refine Tuning of Repository Guidance for Coding Agents

arXiv paper on improving coding agents via repository guidance tuning. Research on optimizing LLM agents for code generation tasks.

Ax Jun He, Deying Yu 23d ago

Sovereign Execution Broker: Enforcing Certificate-Bound Authority in Agentic Control Planes

arXiv paper on secure authority enforcement in autonomous agent control planes. Technical research addressing production agent security and access control.

HN vermaabhishek39 23d ago

Show HN: LoomStack – Helping engineering orgs scale AI-driven development

LoomStack is an orchestration layer for AI-native engineering teams to align autonomous agent execution with organizational processes and policies.

HN vantareed 23d ago

GLM-5.2 is the step change for open agents

Commentary on Zhipu's GLM-5.2 model as advancement for open AI agents, released to coding plan members.

HN renzomad 23d ago

Open-source security auditors for Supabase, Strapi, Hasura and Ollama

Open-source security auditors for backend platforms: Supabase, Strapi, Hasura, Ollama. Local keyless tools confirming leaks via active probes.

HN datadrivenangel 23d ago

The AI Tarpit: Why You Can't Stop Reading Your Code

Essay on why reading code remains essential despite AI coding tools, discussing vibe coding trade-offs and when AI-assisted development works.

HN totetsu 23d ago

King's study finds AI chose nuclear signalling in 95% of simulated crises

King's College London study found LLMs escalated conflicts by choosing nuclear threats in 95% of simulated war game scenarios, raising concerns about AI in military applications.

HN akg268 23d ago

Prompt Preflight – catch vague AI-agent prompts before they burn tokens

Prompt Preflight is a local CLI tool that validates AI agent prompts for specificity before execution, using deterministic rules to catch vague requests and prevent wasted tokens.

HN sriramster 23d ago

Chain-of-Trust with AI

Concept extending chain-of-trust security principles to AI agentic code generation, addressing provenance, auditability, and attestation.

HN leemoore 23d ago

AWS Lambda MicroVMs for isolated execution of user and AI-generated code

AWS Lambda MicroVMs enable VM-level isolation for executing user and AI-generated code with instant launch and state preservation.

HN lobo_tuerto 23d ago

DeepSeek Just Solved AI's Billion Dollar Problem [video]

Video title claiming DeepSeek solved an unspecified AI problem; lacks substantive content.

HN doanbactam 23d ago

Terminal coding agent powered by Kimchi's multi-model orchestration

Terminal CLI coding agent powered by Kimchi multi-model orchestration. AI development assistant with model switching and LLM infrastructure integration.

HN shipfastcli 23d ago

Show HN: ShipFast – Deploy any project to production with one command

One-command deployment tool for projects to production with automatic project type detection. Paid service.

HN john_victor 23d ago

WaterToGo – JavaScript/TS/Python/Rust Codebases to Go

Tool to convert JS/TS/Python/Rust codebases to idiomatic Go using Google Gemini API.

HN robdiorio 23d ago

Show HN: Sonic AI – We made the market's highest-signal audio searchable

Sonic AI makes podcast and earnings call audio searchable by extracting attributed claims and generating cited research briefs using LLM analysis.

HN iBelieve 23d ago

Fallow: Deterministic codebase intelligence for TypeScript and JavaScript

Static analysis tool for JS/TS codebases providing quality metrics, risk assessment, and architecture analysis for humans and agents. Open source analyzer.